pager.c

sqlite数据库的源代码, 例子是在windows上使用,不过可以移值到各种平台上使用

*/
static int pager_delmaster(const char *zMaster){
  int rc;
  int master_open = 0;
  OsFile *master = 0;
  char *zMasterJournal = 0; /* Contents of master journal file */
  i64 nMasterJournal;       /* Size of master journal file */

  /* Open the master journal file exclusively in case some other process
  ** is running this routine also. Not that it makes too much difference.
  */
  rc = sqlite3OsOpenReadOnly(zMaster, &master);
  if( rc!=SQLITE_OK ) goto delmaster_out;
  master_open = 1;
  rc = sqlite3OsFileSize(master, &nMasterJournal);
  if( rc!=SQLITE_OK ) goto delmaster_out;

  if( nMasterJournal>0 ){
    char *zJournal;
    char *zMasterPtr = 0;

    /* Load the entire master journal file into space obtained from
    ** sqliteMalloc() and pointed to by zMasterJournal. 
    */
    zMasterJournal = (char *)sqliteMalloc(nMasterJournal);
    if( !zMasterJournal ){
      rc = SQLITE_NOMEM;
      goto delmaster_out;
    }
    rc = sqlite3OsRead(master, zMasterJournal, nMasterJournal);
    if( rc!=SQLITE_OK ) goto delmaster_out;

    zJournal = zMasterJournal;
    while( (zJournal-zMasterJournal)<nMasterJournal ){
      if( sqlite3OsFileExists(zJournal) ){
        /* One of the journals pointed to by the master journal exists.
        ** Open it and check if it points at the master journal. If
        ** so, return without deleting the master journal file.
        */
        OsFile *journal = 0;
        int c;

        rc = sqlite3OsOpenReadOnly(zJournal, &journal);
        if( rc!=SQLITE_OK ){
          goto delmaster_out;
        }

        rc = readMasterJournal(journal, &zMasterPtr);
        sqlite3OsClose(&journal);
        if( rc!=SQLITE_OK ){
          goto delmaster_out;
        }

        c = zMasterPtr!=0 && strcmp(zMasterPtr, zMaster)==0;
        sqliteFree(zMasterPtr);
        if( c ){
          /* We have a match. Do not delete the master journal file. */
          goto delmaster_out;
        }
      }
      zJournal += (strlen(zJournal)+1);
    }
  }
  
  sqlite3OsDelete(zMaster);

delmaster_out:
  if( zMasterJournal ){
    sqliteFree(zMasterJournal);
  }  
  if( master_open ){
    sqlite3OsClose(&master);
  }
  return rc;
}

/*
** Make every page in the cache agree with what is on disk.  In other words,
** reread the disk to reset the state of the cache.
**
** This routine is called after a rollback in which some of the dirty cache
** pages had never been written out to disk.  We need to roll back the
** cache content and the easiest way to do that is to reread the old content
** back from the disk.
*/
static int pager_reload_cache(Pager *pPager){
  PgHdr *pPg;
  int rc = SQLITE_OK;
  for(pPg=pPager->pAll; pPg; pPg=pPg->pNextAll){
    char zBuf[SQLITE_MAX_PAGE_SIZE];
    if( !pPg->dirty ) continue;
    if( (int)pPg->pgno <= pPager->origDbSize ){
      rc = sqlite3OsSeek(pPager->fd, pPager->pageSize*(i64)(pPg->pgno-1));
      if( rc==SQLITE_OK ){
        rc = sqlite3OsRead(pPager->fd, zBuf, pPager->pageSize);
      }
      TRACE3("REFETCH %d page %d\n", PAGERID(pPager), pPg->pgno);
      if( rc ) break;
      CODEC1(pPager, zBuf, pPg->pgno, 2);
    }else{
      memset(zBuf, 0, pPager->pageSize);
    }
    if( pPg->nRef==0 || memcmp(zBuf, PGHDR_TO_DATA(pPg), pPager->pageSize) ){
      memcpy(PGHDR_TO_DATA(pPg), zBuf, pPager->pageSize);
      if( pPager->xReiniter ){
        pPager->xReiniter(PGHDR_TO_DATA(pPg), pPager->pageSize);
      }else{
        memset(PGHDR_TO_EXTRA(pPg, pPager), 0, pPager->nExtra);
      }
    }
    pPg->needSync = 0;
    pPg->dirty = 0;
#ifdef SQLITE_CHECK_PAGES
    pPg->pageHash = pager_pagehash(pPg);
#endif
  }
  pPager->pDirty = 0;
  return rc;
}

/*
** Truncate the main file of the given pager to the number of pages
** indicated.
*/
static int pager_truncate(Pager *pPager, int nPage){
  assert( pPager->state>=PAGER_EXCLUSIVE );
  return sqlite3OsTruncate(pPager->fd, pPager->pageSize*(i64)nPage);
}

/*
** Playback the journal and thus restore the database file to
** the state it was in before we started making changes.  
**
** The journal file format is as follows: 
**
**  (1)  8 byte prefix.  A copy of aJournalMagic[].
**  (2)  4 byte big-endian integer which is the number of valid page records
**       in the journal.  If this value is 0xffffffff, then compute the
**       number of page records from the journal size.
**  (3)  4 byte big-endian integer which is the initial value for the 
**       sanity checksum.
**  (4)  4 byte integer which is the number of pages to truncate the
**       database to during a rollback.
**  (5)  4 byte integer which is the number of bytes in the master journal
**       name.  The value may be zero (indicate that there is no master
**       journal.)
**  (6)  N bytes of the master journal name.  The name will be nul-terminated
**       and might be shorter than the value read from (5).  If the first byte
**       of the name is \000 then there is no master journal.  The master
**       journal name is stored in UTF-8.
**  (7)  Zero or more pages instances, each as follows:
**        +  4 byte page number.
**        +  pPager->pageSize bytes of data.
**        +  4 byte checksum
**
** When we speak of the journal header, we mean the first 6 items above.
** Each entry in the journal is an instance of the 7th item.
**
** Call the value from the second bullet "nRec".  nRec is the number of
** valid page entries in the journal.  In most cases, you can compute the
** value of nRec from the size of the journal file.  But if a power
** failure occurred while the journal was being written, it could be the
** case that the size of the journal file had already been increased but
** the extra entries had not yet made it safely to disk.  In such a case,
** the value of nRec computed from the file size would be too large.  For
** that reason, we always use the nRec value in the header.
**
** If the nRec value is 0xffffffff it means that nRec should be computed
** from the file size.  This value is used when the user selects the
** no-sync option for the journal.  A power failure could lead to corruption
** in this case.  But for things like temporary table (which will be
** deleted when the power is restored) we don't care.  
**
** If the file opened as the journal file is not a well-formed
** journal file then all pages up to the first corrupted page are rolled
** back (or no pages if the journal header is corrupted). The journal file
** is then deleted and SQLITE_OK returned, just as if no corruption had
** been encountered.
**
** If an I/O or malloc() error occurs, the journal-file is not deleted
** and an error code is returned.
*/
static int pager_playback(Pager *pPager){
  i64 szJ;                 /* Size of the journal file in bytes */
  u32 nRec;                /* Number of Records in the journal */
  int i;                   /* Loop counter */
  Pgno mxPg = 0;           /* Size of the original file in pages */
  int rc;                  /* Result code of a subroutine */
  char *zMaster = 0;       /* Name of master journal file if any */

  /* Figure out how many records are in the journal.  Abort early if
  ** the journal is empty.
  */
  assert( pPager->journalOpen );
  rc = sqlite3OsFileSize(pPager->jfd, &szJ);
  if( rc!=SQLITE_OK ){
    goto end_playback;
  }

  /* Read the master journal name from the journal, if it is present.
  ** If a master journal file name is specified, but the file is not
  ** present on disk, then the journal is not hot and does not need to be
  ** played back.
  */
  rc = readMasterJournal(pPager->jfd, &zMaster);
  assert( rc!=SQLITE_DONE );
  if( rc!=SQLITE_OK || (zMaster && !sqlite3OsFileExists(zMaster)) ){
    sqliteFree(zMaster);
    zMaster = 0;
    if( rc==SQLITE_DONE ) rc = SQLITE_OK;
    goto end_playback;
  }
  sqlite3OsSeek(pPager->jfd, 0);
  pPager->journalOff = 0;

  /* This loop terminates either when the readJournalHdr() call returns
  ** SQLITE_DONE or an IO error occurs. */
  while( 1 ){

    /* Read the next journal header from the journal file.  If there are
    ** not enough bytes left in the journal file for a complete header, or
    ** it is corrupted, then a process must of failed while writing it.
    ** This indicates nothing more needs to be rolled back.
    */
    rc = readJournalHdr(pPager, szJ, &nRec, &mxPg);
    if( rc!=SQLITE_OK ){ 
      if( rc==SQLITE_DONE ){
        rc = SQLITE_OK;
      }
      goto end_playback;
    }

    /* If nRec is 0xffffffff, then this journal was created by a process
    ** working in no-sync mode. This means that the rest of the journal
    ** file consists of pages, there are no more journal headers. Compute
    ** the value of nRec based on this assumption.
    */
    if( nRec==0xffffffff ){
      assert( pPager->journalOff==JOURNAL_HDR_SZ(pPager) );
      nRec = (szJ - JOURNAL_HDR_SZ(pPager))/JOURNAL_PG_SZ(pPager);
    }

    /* If this is the first header read from the journal, truncate the
    ** database file back to it's original size.
    */
    if( pPager->state>=PAGER_EXCLUSIVE && 
        pPager->journalOff==JOURNAL_HDR_SZ(pPager) ){
      assert( pPager->origDbSize==0 || pPager->origDbSize==mxPg );
      rc = pager_truncate(pPager, mxPg);
      if( rc!=SQLITE_OK ){
        goto end_playback;
      }
      pPager->dbSize = mxPg;
    }

    /* Copy original pages out of the journal and back into the database file.
    */
    for(i=0; i<nRec; i++){
      rc = pager_playback_one_page(pPager, pPager->jfd, 1);
      if( rc!=SQLITE_OK ){
        if( rc==SQLITE_DONE ){
          rc = SQLITE_OK;
          pPager->journalOff = szJ;
          break;
        }else{
          goto end_playback;
        }
      }
    }
  }
  /*NOTREACHED*/
  assert( 0 );

end_playback:
  if( rc==SQLITE_OK ){
    rc = pager_unwritelock(pPager);
  }
  if( zMaster ){
    /* If there was a master journal and this routine will return true,
    ** see if it is possible to delete the master journal.
    */
    if( rc==SQLITE_OK ){
      rc = pager_delmaster(zMaster);
    }
    sqliteFree(zMaster);
  }

  /* The Pager.sectorSize variable may have been updated while rolling
  ** back a journal created by a process with a different PAGER_SECTOR_SIZE
  ** value. Reset it to the correct value for this process.
  */
  pPager->sectorSize = PAGER_SECTOR_SIZE;
  return rc;
}

/*
** Playback the statement journal.
**
** This is similar to playing back the transaction journal but with
** a few extra twists.
**
**    (1)  The number of pages in the database file at the start of
**         the statement is stored in pPager->stmtSize, not in the
**         journal file itself.
**
**    (2)  In addition to playing back the statement journal, also
**         playback all pages of the transaction journal beginning
**         at offset pPager->stmtJSize.
*/
static int pager_stmt_playback(Pager *pPager){
  i64 szJ;                 /* Size of the full journal */
  i64 hdrOff;
  int nRec;                /* Number of Records */
  int i;                   /* Loop counter */
  int rc;

  szJ = pPager->journalOff;
#ifndef NDEBUG 
  {
    i64 os_szJ;
    rc = sqlite3OsFileSize(pPager->jfd, &os_szJ);
    if( rc!=SQLITE_OK ) return rc;
    assert( szJ==os_szJ );
  }
#endif

  /* Set hdrOff to be the offset to the first journal header written
  ** this statement transaction, or the end of the file if no journal
  ** header was written.
  */
  hdrOff = pPager->stmtHdrOff;
  assert( pPager->fullSync || !hdrOff );
  if( !hdrOff ){
    hdrOff = szJ;
  }
  
  /* Truncate the database back to its original size.
  */
  if( pPager->state>=PAGER_EXCLUSIVE ){
    rc = pager_truncate(pPager, pPager->stmtSize);
  }
  pPager->dbSize = pPager->stmtSize;

  /* Figure out how many records are in the statement journal.
  */
  assert( pPager->stmtInUse && pPager->journalOpen );
  sqlite3OsSeek(pPager->stfd, 0);
  nRec = pPager->stmtNRec;
  
  /* Copy original pages out of the statement journal and back into the
  ** database file.  Note that the statement journal omits checksums from
  ** each record since power-failure recovery is not important to statement
  ** journals.
  */
  for(i=nRec-1; i>=0; i--){
    rc = pager_playback_one_page(pPager, pPager->stfd, 0);
    assert( rc!=SQLITE_DONE );
    if( rc!=SQLITE_OK ) goto end_stmt_playback;
  }